Simply pytests `subjects_csv` and `sessions_csv` by removing pandas dependency

[CBroz1]

I think the use of pandas here is unjustified. All the operations where it's used can be just more cleanly accomplished with python datatypes: dicts and lists.

Originally posted by @dimitri-yatsenko in datajoint/workflow-session#9 (comment)

writing a sample CSV file with sample data can be made much simpler. Consider simply writing a multiline string into a file.

Originally posted by @dimitri-yatsenko in datajoint/workflow-session#9 (comment)

[ttngu207]

This is reasonable, but I'm not sure about priority.
Pandas is already a dependency for DataJoint, so reimplementing this will not net us a reduction in package dependency.
For code simplification, yes, but certainly will take some time re-implementing (and copying) across the different test suites we have.